188 BYZANTINE CHURCHES CHAP, ix Byzantine construction. Secondly, the set back above the middle string-course on the other walls of the church is an ordinary arrangement in a Byzantine church, and if this were all c the widening? for which Professor Goodyear contended there would be no room for difference of opinion. The ledge formed by that set back may have served to support scaffolding. In the next place, due weight must be given to the distortion which would inevitably occur in Byzantine buildings. They were fabrics of mortar with brick rather than of brick with mortar, and consequently too elastic not to settle to a large extent in the course of erection. Hence is it that no measurements of a Byzantine structure, even on the ground floor, are accurate within more than 5 cm., while above the ground they vary , to a much greater degree, rendering minute measurements quite valueless. Lastly3 as the marble panelling was fitted after the completion of the body of the building, it had to be adapted to any divergence that had previously occurred in the settling or the walls or the spreading of the vaults. The marble panelling, it should also be observed, is here cut to the diagonal at one angle, and not at the other. Apart from the set back.of the masonry at the middle string-course, this church, therefore, supplies no evidence for an intentional widening of . the structure from the ground upwards. Any further widening than that at the middle string-course was accidental, due to the nature of the materials employed, not to the device of the builder, and was allowed by the architect because unavoidable. Such irregularities are inherent in the Byzantine methods of building.